Hardware Acceleration

Modeling of plasma dynamics is challenging due to the wide range of spatial and temporal scales involved in the complex behavior of the system. Because the system behavior is mainly encompassed in the coupling of the system elements rather than the components that make up the system, describing its complexity requires self-consistent coupling between the elements.

Prof. Ilie is an integral member of the Space Weather Modeling Framework (SWMF) development team, a multi-institutional team led by the University of Michigan. SWMF is an excellent tool for this purpose since it combines a global magnetospheric solution from the BATS-R-US code with regional solutions of several specific regions in geospace. However, the BATS-R-US model is the most expensive domain within an SWMF simulation, which makes high fidelity simulations computationally expensive. This resource limitation is the primary reason many unanswered questions in magnetospheric physics (and beyond) remain unexplored.

Porting numerical models to GPUs, not only increases the productivity and reduces the simulation cost by reducing running time, but also helps with the energy consumption. In state-of-the-art CPU clusters, energy consumption is huge and currently it is not in the range of budget for scientific community or in general for a state even. Therefore, currently it is not possible to explore certain research areas, which remain uncharted territories full of potential for breakthrough science, unless we use GPUs and alike accelerators.